Bank Marketing Data

Amaya Owens, Paige Hill, Lauren McIntosh

Background Information

  • Data was collected from a Portuguese bank to analyze the effectiveness of their marketing campaign
  • Collected from 2008 to 2013, so it is affected by the financial crisis
  • Selling long-term deposits
  • Previous research only yielded good results when using call duration
  • Allows them to focus resources on customers most likely to subscribe

Dataset Description

  • The researchers began with 150 features and selected 17 of the most relevant features
  • 17 columns and 45,211 rows
  • Features: demographics, financial info, campaign details
  • Some “unknown” values in contact method and previous outcomes

Basic Information

Var Data type
Age int
Job obj
Martial obj
Education obj
Default obj
Balance int
Var Data Type
Housing obj
Loan obj
Contact obj
Day int
Month obj
Duration int
Var Data Type
Campaign int
PDays int
Previous int
POutcome obj
y obj

Statistical Measures

Variable Age (years) Balance (€) Campaign (contacts) Duration (sec)
Minimum 18 -8,019 1 0
Median 39 448 2 180
Maximum 95 102,127 63 4,918
Mean 40.9 1,362 2.8 258
Std Dev 10.6 3,044 3.1 258

Demographics & Features

Category Most Common Count % Total
Job Blue-collar 9,732 21.5%
Education Secondary 22,472 49.7%
Marital Married 27,214 60.2%
Housing Loan Yes 25,130 55.6%
Personal Loan No 37,967 84.0%
Contact Month May 15,629 34.6%

Key Statistics:

  • Age: 18-95 years (mean: 40.9)
  • Balance: 7.9% have a negative balance
  • Campaign: Most contacted 1-2 times
  • Duration: 14.8% had 0-second calls
  • Previous contact: 81.7% never contacted before
  • Target variable: 11.7% subscribed

Subscriptions

Education Level

Age

Frequency of Contact

Outcome of Previous Campaigns

Questions

  • What are the key demographic and financial profiles of clients who subscribe to a term deposit compared to those who do not?
  • How does the strategy of the marketing campaign itself (specifically the timing and frequency of contact) correlate with the success rate of subscriptions?
  • To what extent does a client’s past interaction with the bank, including previous campaign outcomes, predict their likelihood of subscribing for the current campaign?